Goal Reasoning concerns actors that deliberate about their own goals, and learning is often applied to improve performance of such actors. A recent model of Goal Reasoning proposed by Roberts et al. lacks a simplified working example with an explanation of how the model can incorporate learning. We describe a thought experiment modeling the decision making of floor-cleaning robots, which are simple goal reasoning actors. We observe that such robots often have a less-than-optimal action policy that, though sufficient for cleaning in the limit, could be improved through learning to reduce the total cleaning time or reduce energy usage. We start with by modeling the straightforward policy for a simple robot named Vacuous. We then ponder a hypothetical robot, named KleeneStar, which optimally cleans in the fastest time possible. Finally, we describe a model for an improved robot, named Vakleene, which iteratively reduces its cleaning time through learning. While these examples are only a thought experiment, this work provides a model of a simplistic goal reasoning process that can learn.